Bank Abnormal Behaviour Recognition Technology Based on Deep Learning
To resolve the problem of low recognition accuracy of existing bank abnormal behavior recognition methods, a bank abnormal behavior recognition method based on deep learning is proposed. The static background of a bank surveillance video image is acquired by the Mixture-of-Gaussians (MoG) model. The static background is extracted by background subtraction, and the foreground image is filtered to make the moving human target as clear as possible. The video frame image is divided into blocks to obtain the motion effect map of the foreground area, and the motion effect map features of each space-time block are extracted. After obtaining the depth features of moving objects, the sparse reconstruction method is used to introduce
a coefficient learning dictionary and a sparse coding vector to judge whether the behavior is normal or not according to the sparse reconstruction cost. The experimental results show that compared with traditional methods, the proposed method based on deep learning is more accurate and has a higher application value.
Keywords: MoG model, Background subtraction, Feature extraction, Coefficient reconstruction, Abnormal behavior recognition
Y. Du, "Bank Abnormal Behaviour Recognition Technology Based on Deep Learning", Engineering
Intelligent Systems, vol. 29 no. 4, pp. 235-239, 2021.